# SaferChatAI — AI Moderation That Understands Context

AutoMod catches keywords. SaferChatAI catches meaning.

Harassment, hate speech, and toxic behaviour rarely come as exact slurs. They hide in typos, coded language, and "technically not against the rules" phrasing. SaferChatAI uses AI to understand **what someone actually means** — not just what words they used.

---

## The Problem

> "Fuk off, your setup is embarrassing lmao, typical behaviour from people like you"

AutoMod sees no keyword match. **SaferChatAI flags it** — personal attack + targeting, typo or not.

> "not being racist but statistically those communities are just more aggressive, it's just facts lol"

AutoMod sees polite language, no slurs. **SaferChatAI flags it** — coded hate speech with pseudo-statistical framing.

---

## What You See

Every flagged message lands in your mod channel as a detailed report:

- **The full conversation thread** so you understand context at a glance
- **Who said it** with author info
- **Why it was flagged** with a clear AI explanation
- **One-click buttons:** Delete · Warn · Timeout · False Positive · Ban

No log-diving. No guesswork. Your mod team reviews and decides.

---

## Features

🛡️ **No auto-bans. Ever.** The AI flags — your mods decide. Always human control.

🔍 **Context-aware detection** — Understands conversation flow, not just individual messages. Catches what keyword filters miss.

💙 **Crisis detection** — When self-harm language is detected, the user receives a private, compassionate message with crisis resources. Your mod team gets an urgent alert.

⚡ **Custom word filter** — Build your own blocklist. Blocked words are deleted instantly before anyone sees them.

🔗 **Link protection** — 5 configurable modes with scam detection. Block suspicious links, phishing attempts, and known malicious domains.

📊 **Moderation dashboard** — 30-day overview: incidents by category, top offenders, false positive rate, action history per user.

🎯 **Learns from your team** — Every time your mods mark a false positive, SaferChatAI adapts. It calibrates to your server's culture over time.

🌍 **Multi-language support** — Set your server's primary language so the AI evaluates non-English messages correctly.

⚙️ **Per-category sensitivity** — Set each detection category independently to Strict, Standard, Relaxed, or Off. sexual/minors is always active.

---

## Detection Categories

Harassment · Hate Speech · Sexual Content · Sexual/Minors *(always on)* · Self-Harm · Violence · Illicit Activity · Scam Links

---

## Commands

| Command | Who | What |
|---------|-----|------|
| `/config` | Admin | Main setup: mod channel, auto-delete, link protection, word filter, exempt roles/channels |
| `/advanced` | Admin | Per-category sensitivity (Strict / Standard / Relaxed / Off) |
| `/mod history @user` | Mod | Incident & action history for a user |
| `/mod warn @user` | Mod | Send an official warning via DM |
| `/statistics` | Mod | 30-day moderation dashboard |
| `/help` | Everyone | Commands and setup guide |

---

## Setup in 2 Minutes

1. Add SaferChatAI to your server
2. Run `/config` — pick your mod channel
3. Done. Monitoring starts immediately.

Optional: Tune sensitivity per category with `/advanced`, add exempt channels/roles, configure link protection and word filter.

---

## Works With Your Existing Setup

SaferChatAI doesn't replace AutoMod or MEE6 — it's the context layer on top. Your existing keyword rules keep working. SaferChatAI catches everything they miss.

**Free during beta · No credit card · No hidden limits**

🔗 [Website](https://saferchat.org) · [Support Server](https://discord.gg/DnXtuaeK2W) · [Privacy Policy](https://saferchat.org/legal/privacy.html) · [Terms of Service](https://saferchat.org/legal/tos.html)